The default mental model for scaling a hard-tech company has been, for a long time, what you might call the integrated giant model. You raise a billion dollars, build your own fab, hire your own operators, write off your own depreciation, and bear all the operational and capital risk yourself. This is how Intel did it. It's how Tesla did it for the Gigafactory. It's how the Japanese keiretsu did it. It works, sometimes spectacularly, but it's also slow, expensive, and brittle in a particular way: any miss on demand forecasting, any technology shift, any change in market geography becomes very expensive very quickly when the entire balance sheet is committed to a single physical footprint.

I think this model is increasingly the wrong one for a substrate company in 2026, and the reason is that there's a better one available now that wasn't available in the era when the integrated giants got built.

The better model is to compose your manufacturing strategy out of contract relationships with facilities that already exist. The American industrial base, while diminished, didn't vanish. It contracted, repurposed, and dispersed. A lot of the capacity that used to make commodity products at scale — film, fibers, specialty chemicals, coated webs, precision metals — got repurposed for smaller production runs and is now available for use by anyone who can put a viable specification in front of it. The depreciation lives on someone else's balance sheet. The operators are already trained. The yield curve has already been climbed.

For us, the canonical example is Kodak Park in Rochester. Kodak used to coat film for the entire world. The same equipment that made Kodachrome — color film consistent to about one part in ten thousand across enormous webs of substrate— is now available for contract coating. We use it. We don't own it, and we don't need to. The capital we would have spent building a coating line went into the polymer chemistry instead, and that's a much better use of dollars fora company at our stage.

This isn't just a cost-savings argument. It's a structural one. A contract manufacturing model lets you decouple the chemistry from the geometry. If you build your own line, you've made a permanent bet on a particular substrate width, a particular coating speed, a particular roll diameter, a particular curing chemistry. If the application surface above you shifts — and over a decade or two, with a platform technology serving many industries, it almost certainly will — you're stuck. A contract model gives you the option to shift, to add a different line for a different geometry, to scale capacity in months rather than years. It's more flexible.

It's also more resilient. A single fab is a single point of failure. Multiple facilities, in different regions, run by different operators, on different equipment, supplying different customer bases, is a much more robust supply chain. If you wanted to take down a fully integrated manufacturer, you'd attack their one factory. If you wanted to take down a distributed manufacturer, you'd have to attack five facilities owned by five different companies in five different states, and you'd still only get a fraction of the capacity. From a national security standpoint, the defense customers we work with care a lot about this kind of resilience, and rightly so.

There's a second sense of "distributed" that I think matters here,which is geographic. The customer base for a platform technology like ours isn't located in any one place. Fuel cell customer sare in Detroit. Electrolyzer customers are in the chemical clusters along the Gulf Coast and in the hydrogen-using industrial bases of the upper Midwest. Flow battery customers are mostly along the West Coast and in Texas. Defense customers are spread across Huntsville, Norfolk, Aberdeen, and Dayton. A single coating line in Rochester serves a national customer base by air freight, but a longer-term strategy probably involves additional lines closer to the customer geographies for cost and time reasons.

The traditional argument against contract manufacturing is that you give up control of your manufacturing know-how. I think this is mostly a holdover from a different era. The know-how that matters in our category is the polymer chemistry and the coating recipe. The know-how that matters in our partners' facilities is how to run a coating line at industrial scale, which is operational discipline that took Kodak decades to build and that we couldn't replicate even if we tried. The two kinds of know-how are complementary, not competitive. We bring the chemistry; they bring the operational depth. It's a good trade.

There's a less-discussed reason to like this model, which is that it's well-matched to the capital environment for hard-tech companies in 2026. The traditional integrated model required a single very large capital raise to build the fab, which made the financing pattern lumpy and high-stakes. The contract model lets you spend capital incrementally — pay for production as you sell, scale capacity in proportion to demand, avoid the big bet on a single physical footprint. This is a much better match for the way venture and growth capital actually work today, and it means a founder doesn't have to take the kind of binary bet that the previous generation of substrate companies had to take.

I'd predict that more of the next decade of hard-tech companies are going tolook like this. Not all of them — some categories genuinely require dedicated factories, and the integrated model will still make sense for those — but a lot of them. The American industrial base has surface area waiting to be put back to work, and the founders who figure out how to compose strategies around that surface area are going to be much faster, cheaper, and more resilient than the founders who try to build everything from scratch.

The reindustrialization story, in practice, isn't going to look like building hundreds of new vertically integrated mega-factories. It's going to look like a network of specialized facilities — some old, some new, some retrofit, some greenfield — operated by different companies and connected by chemistry and defensible IP rather than by ownership. That's what we're doing. I think that's most of what works in this era.

‍